Manual
Basic manual for using the platform.
Quick start (3-5 minutes)
Everything you need to start and run workloads.
- 1
Create account and verify email.
- 2
Add a payment method in Billing.
- 3
Go to Launch, pick a plan, launch an instance.
- 4
Open Jupyter or SSH from the instance card.
- 5
Save results and terminate the instance.
Save time
Keep reusable notebooks so you can launch faster next time.
Keep your work
Sync /workspace with Google Drive from Settings when needed.
Control costs
Check Billing regularly and terminate instances when done.
Visual checklist
Launch in minutes
Pick a plan and pricing mode, then confirm.
Work in Jupyter
Open Jupyter or SSH and run your workflow.
Save results
Download files or sync with Drive.
Finish and save costs
Terminate the instance when done.
Quick index
User manual
Everything you need to start and run workloads.
Account and access
- Sign up with email and password.
- Verify email to activate the account.
- Login with persistent session if you use "Remember me".
- Profile and security settings in Settings.
Payments and billing
- Add, update, or remove cards in Billing.
- Charges are per minute and grouped automatically.
- If a payment fails, new launches are blocked until you update the card.
- Balance shows Open, Paid, and Failed.
Plans and pricing
- Plans S/M/L/XL: bigger plan means more VRAM and RAM.
- Custom plan to adjust resources and price limits.
- Modes: Standard (stable), Night (cheaper), Spot (can be interrupted).
- Billed per minute based on the hourly price.
What to pick
- S: tests, light notebooks, prototypes.
- M: medium models or multiple sessions.
- L/XL: large training or heavy workloads.
- Custom: set minimum VRAM and adjust CPU/RAM if you need specifics.
Launch an instance
Go to /launch.
Pick a plan or custom mode.
Select pricing mode.
Confirm and launch.
Wait for Provisioning → Running.
Manage instances
- Statuses: Provisioning, Running, Stopping, Terminated.
- Stop or terminate when done.
- Cards show costs, time, and access links.
Access to GPU (Jupyter / SSH)
Go to Settings -> SSH keys, choose a clear device name (for example: windows-wsl, macbook, linux-laptop), and click "Generate and save for this device". Clodei stores only the public key; the private key is generated in your browser.
Download the private key and note the displayed SHA256 fingerprint. The fingerprint helps you confirm you are using the same key later if SSH returns "Permission denied (publickey)".
If you use WSL and downloaded the key in Windows, copy it into ~/.ssh and restrict permissions. OpenSSH on WSL commonly rejects keys used directly from /mnt/c/... because permissions are too open.
mkdir -p ~/.ssh
cp /mnt/c/Users/<your-user>/Downloads/<your-key> ~/.ssh/<your-key>
chmod 600 ~/.ssh/<your-key>In /launch, enable SSH. Launch automatically preselects the default key for this device (or the last one you selected). Confirm the fingerprint before launching.
From the instance details page, use the SSH command shown by Clodei. It should include -i with your private key path. Replace <your-key> and <ssh-host> if you copy the example manually.
ssh -o ServerAliveInterval=30 -o ServerAliveCountMax=3 -o ProxyCommand='cloudflared access ssh --hostname %h' -i ~/.ssh/<your-key> clodei@<ssh-host>If you see "Permission denied (publickey)", the instance usually started with a different authorized key. Compare the instance fingerprint with your local key fingerprint and relaunch selecting the correct key in Launch.
- Jupyter is preinstalled and opens with token.
- SSH is available if the plan includes it and requires a public/private key pair.
- Instance SSH username is fixed: clodei.
- The SSH command shown by Clodei includes -i (private key path).
Data and storage
- Save results before ending the instance.
- Storage is removed when the instance ends.
- Use Drive or external backups for persistence.
Costs and control
- Balance and charges in Billing.
- Charge history with Stripe reference.
- Internal invoices downloadable in Invoices.
Common issues
- Cannot launch: check payment method and balance.
- No capacity: wait or reduce requirements.
- Spot interrupted: use checkpoints.
- Jupyter not opening: check token and Running status.
Support
- Go to /support to contact us.
- Include instance ID and approximate time.

